L^p-form of the KNRS conjecture
Pith reviewed 2026-06-30 05:33 UTC · model grok-4.3
The pith
For p at least binom(n,2)/m every rho-locally dense graphon satisfies the L^p lower bound on homomorphic copies of any fixed graph F.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
If p is at least binom(n,2)/m then for every rho-locally dense graphon W, t(F, W to the circle p) is at least rho to the pm, equivalently the L^p norm of W_F is at least rho to the e(F). The proof proceeds by Holder uniformization over vertex relabellings. A comparison principle with edge-transitive KNRS supergraphs gives sharper exponents in some cases, and positive-semidefinite methods show that Sidorenko-good graphs are closed under uniform theta-subdivisions, that the non-uniform theta theorem lifts after removing the parity restriction, and that uniform theta-subdivisions of KNRS graphs are regular-KNRS.
What carries the argument
Holder uniformization over vertex relabellings, which converts the local-density hypothesis on W into the desired L^p lower bound on the edge-product function W_F.
If this is right
- Every graph with at least one edge satisfies the stated L^p form of the KNRS conjecture.
- Sharper exponents hold whenever F embeds into an edge-transitive KNRS graph.
- Sidorenko-good graphs remain Sidorenko-good under arbitrary uniform theta-subdivisions.
- Uniform theta-subdivisions of KNRS graphs are regular-KNRS.
- The non-uniform theta theorem admits a Sidorenko-good lift after removing the parity restriction under the same divisibility assumptions.
Where Pith is reading between the lines
- The L^p bound may serve as an intermediate step toward the full KNRS conjecture if it can be upgraded by letting p grow or by combining it with other density-increment arguments.
- Edge-transitive graphs appear to play a distinguished role because the comparison principle improves the exponent precisely when such a supergraph exists.
- The subdivision closure results suggest that both the Sidorenko and KNRS properties are stable under uniform theta-subdivision operations.
Load-bearing premise
A Holder uniformization step over vertex relabellings exists that converts the local-density hypothesis into the L^p lower bound under the needed uniformity and measurability conditions on the graphon.
What would settle it
A concrete graph F, exponent p equal to binom(n,2)/m, and rho-locally dense graphon W for which the L^p norm of the edge-product W_F falls strictly below rho to the power e(F).
read the original abstract
The Kohayakawa--Nagle--R\"odl--Schacht conjecture predicts that locally dense graphs contain, asymptotically, at least as many homomorphic copies of any fixed graph as the random graph of the same edge density. We prove that every graph with at least one edge satisfies a natural $L^p$ relaxation of this conjecture in the graphon setting. More precisely, let $F$ be a graph with $m>0$ edges, and let $n$ be the number of non-isolated vertices of $F$. If $$ p\ge \binom {n}{2}/m, $$ then for every $\rho$-locally dense graphon $W$, $$ t(F,W^{\circ p})\ge \rho^{pm}. $$ Equivalently, if $$ W_F(\mathbf x)=\prod_{ij\in E(F)}W(x_i,x_j), $$ then $$ \|W_F\|_{L^p}\ge \rho^{e(F)}. $$ The proof is based on a H\"older uniformization over vertex relabellings, in the spirit of Conlon--Lee. We also prove a more general comparison principle with edge-transitive KNRS supergraphs, yielding sharper exponents whenever $F$ embeds into an edge-transitive KNRS graph. Finally, positive-semidefinite methods give theta-subdivision results: Sidorenko-good graphs are closed under arbitrary uniform theta-subdivisions; the non-uniform theta theorem of Im--Li--Liu admits a Sidorenko-good lift, under the same divisibility assumptions, after removing the parity restriction; and uniform theta-subdivisions of KNRS graphs are regular-KNRS.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proves an L^p relaxation of the KNRS conjecture in the graphon setting: for a graph F with m>0 edges and n non-isolated vertices, if p ≥ binom(n,2)/m then every ρ-locally dense graphon W satisfies t(F, W^{∘p}) ≥ ρ^{p m}, or equivalently ||W_F||_{L^p} ≥ ρ^{e(F)}. The proof uses Hölder uniformization over vertex relabellings in the spirit of Conlon-Lee. It also establishes a comparison principle yielding sharper exponents for F embeddable in edge-transitive KNRS graphs, and positive-semidefinite results on theta-subdivisions: Sidorenko-good graphs are closed under uniform theta-subdivisions, the Im-Li-Liu non-uniform theta theorem lifts to Sidorenko-good graphs without parity restriction under divisibility assumptions, and uniform theta-subdivisions of KNRS graphs are regular-KNRS.
Significance. If the central L^p inequality holds, the result supplies a natural and potentially useful relaxation of the full KNRS conjecture, together with a technique (Hölder uniformization) that may extend to stronger statements. The comparison principle and the subdivision closure results are of independent interest and strengthen the known landscape around Sidorenko and KNRS properties. The manuscript gives a clean, parameter-free statement of the exponent threshold and credits the Conlon-Lee inspiration explicitly.
major comments (1)
- [Proof of the main theorem (Hölder uniformization step)] The proof of the main L^p inequality (the statement in the abstract and presumably Theorem 1.1) proceeds by averaging t(F, W^σ) over vertex relabellings σ and applying Hölder to pass from the ρ-local-density hypothesis on each W^σ to the desired bound on ||W_F||_{L^p}. This step requires that the map σ ↦ W^σ preserve measurability and that the local-density lower bound transfers to the averaged object; the manuscript invokes the uniformization 'in the spirit of Conlon-Lee' but does not record the σ-algebra on the space of relabellings or the precise integrability class of W needed for the integrals defining t(F, W^{∘p}) to be well-defined and interchangeable with the supremum over σ.
minor comments (1)
- [Abstract] In the abstract the notation W^{∘p} and W_F(𝐱) is used before the definitions that appear later in the text; a brief parenthetical reminder would improve readability.
Simulated Author's Rebuttal
We thank the referee for the careful reading and for highlighting the need for greater measure-theoretic precision in the Hölder uniformization argument. We address the single major comment below and will incorporate the requested clarifications in the revised manuscript.
read point-by-point responses
-
Referee: The proof of the main L^p inequality (the statement in the abstract and presumably Theorem 1.1) proceeds by averaging t(F, W^σ) over vertex relabellings σ and applying Hölder to pass from the ρ-local-density hypothesis on each W^σ to the desired bound on ||W_F||_{L^p}. This step requires that the map σ ↦ W^σ preserve measurability and that the local-density lower bound transfers to the averaged object; the manuscript invokes the uniformization 'in the spirit of Conlon-Lee' but does not record the σ-algebra on the space of relabellings or the precise integrability class of W needed for the integrals defining t(F, W^{∘p}) to be well-defined and interchangeable with the supremum over σ.
Authors: We agree that the current write-up leaves the underlying measure space implicit. In the revision we will add a short subsection (or appendix paragraph) that (i) models the space of relabellings as the compact group of measure-preserving transformations of [0,1] equipped with its Haar probability measure, (ii) verifies that σ ↦ W^σ is measurable with respect to the product σ-algebra and preserves the L^1 (or L^∞) class of W, (iii) shows that the ρ-local-density lower bound passes to the averaged graphon by Fubini, and (iv) justifies the interchange of the integral defining t(F,W^{∘p}) with the supremum over σ via the dominated-convergence theorem under the standing assumption that W is bounded (as is standard for graphons). These additions will make the Hölder-uniformization step fully rigorous while remaining in the spirit of the Conlon–Lee argument. revision: yes
Circularity Check
No circularity; derivation proceeds via Holder on averaged relabellings
full rationale
The paper states that the L^p lower bound follows from the rho-local-density hypothesis by averaging t(F, W^sigma) over vertex relabellings sigma and applying Holder, in the spirit of Conlon-Lee. This is an inequality step from the given assumption to the claimed bound on ||W_F||_L^p, with no reduction of the left-hand side to fitted constants, no self-definitional equivalence, and no load-bearing self-citation for the central claim. The reference to Conlon-Lee supplies the averaging technique but does not make the present result equivalent to prior inputs by construction. No steps match the enumerated circularity patterns.
Axiom & Free-Parameter Ledger
axioms (2)
- standard math Holder's inequality applies to the product functions arising from graphons
- domain assumption Graphons are measurable functions [0,1]^2 → [0,1] and homomorphism densities are well-defined integrals
Forward citations
Cited by 1 Pith paper
-
Tensor Amplification and Spectral Transfer for Sidorenko-Type Inequalities
Tensor-amplification framework proves equality regularization and spectral equivalence for C-Sidorenko graphs in admissible graphon classes.
Reference graph
Works this paper leans on
-
[1]
Counting subgraphs in locally dense graphs.Israel Journal of Mathematics, 2025
Domagoj Bradaˇ c, Benny Sudakov, and Yuval Wigderson. Counting subgraphs in locally dense graphs.Israel Journal of Mathematics, 2025. To appear
2025
-
[2]
Kohayakawa–Nagle–R¨ odl–Schacht conjecture for subdivisions.arXiv preprint arXiv:2407.10861, 2024
Hao Chen, Yupeng Lin, and Jie Ma. Kohayakawa–Nagle–R¨ odl–Schacht conjecture for subdivisions.arXiv preprint arXiv:2407.10861, 2024
-
[3]
Sidorenko’s conjecture for blow-ups.Discrete Anal- ysis, pages Paper No
David Conlon and Joonkyung Lee. Sidorenko’s conjecture for blow-ups.Discrete Anal- ysis, pages Paper No. 2, 13 pp., 2021. doi: 10.19086/da.21472
-
[4]
Some advances on Sidorenko’s conjecture.Journal of the London Mathematical Society, 98(3):593–608,
David Conlon, Jeong Han Kim, Choongbum Lee, and Joonkyung Lee. Some advances on Sidorenko’s conjecture.Journal of the London Mathematical Society, 98(3):593–608,
-
[5]
doi: 10.1112/jlms.12142
-
[6]
A limit theorem in graph theory.Studia Scientiarum Mathematicarum Hungarica, 1:51–57, 1966
Paul Erd˝ os and Mikl´ os Simonovits. A limit theorem in graph theory.Studia Scientiarum Mathematicarum Hungarica, 1:51–57, 1966
1966
-
[7]
Seonghyuk Im, Ruonan Li, and Hong Liu. Sidorenko’s conjecture for subdivisions and theta substitutions.Combinatorics, Probability and Computing, 35(2):269–279, 2026. doi: 10.1017/S0963548325100242
-
[8]
Forcing Graphs to Be Forcing.arXiv preprint arXiv:2412.12904, 2024
Aldo Kiem, Olaf Parczyk, and Christoph Spiegel. Forcing Graphs to Be Forcing.arXiv preprint arXiv:2412.12904, 2024
-
[9]
Yoshiharu Kohayakawa, Brendan Nagle, Vojtˇ ech R¨ odl, and Mathias Schacht. Weak hypergraph regularity and linear hypergraphs.Journal of Combinatorial Theory, Series B, 100(2):151–160, 2010. doi: 10.1016/j.jctb.2009.05.005
-
[10]
On some graph densities in locally dense graphs.Random Structures & Algorithms, 58(2):322–344, 2021
Joonkyung Lee. On some graph densities in locally dense graphs.Random Structures & Algorithms, 58(2):322–344, 2021. doi: 10.1002/rsa.20974
-
[11]
Counting odd cycles in locally dense graphs.Journal of Combinatorial Theory, Series B, 105:1–5, 2014
Christian Reiher. Counting odd cycles in locally dense graphs.Journal of Combinatorial Theory, Series B, 105:1–5, 2014. doi: 10.1016/j.jctb.2013.12.002. 17
-
[12]
Alexander Sidorenko. A correlation inequality for bipartite graphs.Graphs and Com- binatorics, 9(2):201–204, 1993. doi: 10.1007/BF02988307
-
[13]
Alexander Sidorenko. Inequalities for Doubly Nonnegative Functions.The Electronic Journal of Combinatorics, 28(1):#P1.32, 2021. doi: 10.37236/8947. 18 A Proof of the relations between the notions Proof of Theorem 1.4.For (1), letWbe aρ-locally dense graphon. TakingS= [0,1] in the definition of local density gives t(K2, W) = Z [0,1]2 W(x, y) dxdy≥ρ. IfGis ...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.